1 |
On Fri, 2006-07-07 at 02:31 +0200, Luca Barbato wrote: |
2 |
> The more I think about the issue and the more I like the complete |
3 |
> profiles for amd64 more than the other solutions. |
4 |
|
5 |
I don't even *want* to think of what this would be for x86. |
6 |
|
7 |
These are what I can think of, so far, with regards to different support |
8 |
on different chips. |
9 |
|
10 |
x86 (everything) |
11 |
i586 (everything i586-compatible) |
12 |
i586 + mmx (pentium-mmx) |
13 |
i686 (everything i686-compatible) |
14 |
i686 + mmx (pentium2+, athlon+) |
15 |
i686 + mmx + sse (pentium3+, athlon-xp+) |
16 |
i686 + mmx + sse + sse2 (pentium4+, athlon64+, opteron+) |
17 |
i686 + mmx + see + sse2 + sse3 (some pentium4, some athlon64, some |
18 |
opteron) |
19 |
i686 + mmx + 3dnow (athlon+) |
20 |
i686 + mmx + 3dnow + sse (athlon-xp+) |
21 |
i686 + mmx + 3dnow + sse + sse2 (athlon64+, opteron+) |
22 |
i686 + mmx + 3dnow + sse + sse2 + sse3 (some athlon64, some opteron) |
23 |
|
24 |
Now, some of those aren't able to be turned on solely via -march. |
25 |
|
26 |
I'm not arguing for or against this, since I haven't bothered to read |
27 |
the entire thread at the moment. I just wanted to point out that we |
28 |
would likely end up with 12 sub-profiles for all of our profiles to |
29 |
accomplish this. Even if we only started this going forward, x86 has a |
30 |
few profiles considered "supported" by Release Engineering that would |
31 |
need adjustment... |
32 |
|
33 |
x86/2006.1/desktop |
34 |
x86/no-nptl |
35 |
x86/no-nptl/2.4 |
36 |
|
37 |
This means it is now 36 profiles to support, if we dropped support on |
38 |
all profiles except for the new ones. Without having any sort of |
39 |
multiple inheritance available, this is really unmanageable. |
40 |
|
41 |
-- |
42 |
Chris Gianelloni |
43 |
Release Engineering - Strategic Lead |
44 |
x86 Architecture Team |
45 |
Games - Developer |
46 |
Gentoo Linux |